Navigating the digital landscape in densely populated locations characterized by heightened diversity poses additional challenges to voter persuasion. Campaigns often run into the problem of persuading populations whose first language isn’t English to vote, given existing language barriers and the difficulty in identifying and targeting these populations in their preferred language. To mobilize hard-to-reach voters across the city, The New York City Campaign Finance Board, an independent agency charged with spreading voting information, chose to target immigrant voters. DSPolitical partnered with Fenton and INTRVL to reach these voters through a cross-channel media campaign focused on education, mobilization, and building support for NYC Votes, an initiative to boost participation among voters and candidates alike to ensure NYC elected officials reflect the diversity of the city and the needs of the communities. Our approach to refining audiences with experimental design and language modeling resulted in statistically significant lifts in favorability toward NYC Votes.

Experimental Design & Measurement Modeling 

The digital experiment was built by randomizing the immigrant audiences into treatment and control groups. Due to the high number of multi-unit apartment complexes in NYC, randomization was performed at the zip code level, which minimizes waste and enables us to target voters with an additional identifier. Additionally, demographic characteristics were incorporated into the randomization through a blocking process to ensure adequate balance between treatment and control groups. We implemented causal inference modeling to control for survey response bias, address any remaining imbalance between treatment and control groups, and measure the effect of the media campaign across the audience as a whole and within subgroups.


  • Max Opportunity Targeting Surveys: In order to build refined audiences, surveys, and models were run to determine the best language for each member of the audience, whether it be Spanish, Cantonese, Mandarin, or English. This wave aims to identify what language the audience prefers to consume online content in so that media can be in their preferred language alongside English.
  • Post-Campaign Surveys: The purpose of this wave was to measure public opinion on NYC Votes.

We leveraged pre- and post-campaign surveys to assess the impact of media programs and language modeling. The Post-Campaign Surveys provided our client with invaluable insights into more effective voter outreach strategies and optimal media spending allocation and revealed the potential impact of running media campaigns in the audience’s preferred language. The surveys yielded statistically significant results for immigrants moving up the favorability ladder overall based on their responses during the pre-campaign surveys.

Post-Survey Results

  • Top 1 – Very Positive Average Lift: +2.0pp
  • Top 2 – Very & Somewhat Positive Average Lift: +2.1pp
  • Top 3 – Very & Somewhat Positive, Unsure Average Lift: +4.1pp

In the increases of favorability shown, respondents who were initially unsure about NYC Votes before the media campaign exhibited the highest lift (4.1pp). The results indicate that language modeling had a particularly impactful change in perception among this demographic. Overall, there was an increase in favorability toward NYC Votes across all surveyed groups.

At DSPolitical, we possess the necessary tools and expertise to execute successful media campaigns, even in the most diverse environments, ensuring the effective mobilization of voters.